Recognition and drinking behaviour analysis of individual pigs based on machine vision

Abstract

Abstract Water consumption by individual pigs can be an interesting indicator of their health. A method using machine vision is proposed to (a) recognise the presence of an individual pig within the drinking zone and (b) analyse the vision images to determine if the pig is drinking. First, isolation of an individual pig within the drinking zone is extracted from the topview of the set of video sequences for group-housed pigs. Next, the distance between the individual pig and the drink nipple is calculated and used to determine whether the individual pig is in contact with the drink nipple. If yes, the colour moments, area, perimeter and other features of the pig are extracted. Then the features are normalised. The individual pig is recognised by computing the Euclidean distance between the pig and the standard sample. The contact time between an individual pig and the drink nipple is used to determine whether the pig is drinking. The pigsty contains 7 pigs and is monitored in real-time, and 140 video clips containing images of the individual pigs while drinking are captured. The correct (drinking) recognition rate for individual pigs is 90.7%. Our method differs from traditional methods in that it avoids any disturbance to the pigs, and it can be used for the recognition of individual pigs within a stress-free environment. Our results can provide a reference point and direction for exploration of other behaviours of topview group-housed pigs.